Telugu Large Language Model Project

Corpus Database

• Maintains curated Telugu documents of different genres

• PostgreSQL Server

• Foundation for everything else

Web Application

• Provides a workflow system for curating Telugu documents

• Role-based access control

Technical Volunteers for Developing the Web Application

• React.js

• Django (secure data and content management) vs Node.js (real-time data processing and scalability under heavy traffic)

• Software Testing

• PostgreSQL Administration

• Linux Server Administration

• Doccano Server Administration

Telugu Linguists

• Annotation using the Doccano Server

• Coarse-grained vs fine-grained POS tags

• Paraphrasing problem: binary classification (yes/no)

• Textual entailment: binary (yes/no) vs. ternary (yes/no/neutral) classification
  

• Annotated data for question answering, text summarization, and text classification

Demos

• Telugu BERT Language Model

• Telugu Comprehension Model (Question Answering)

• Telugu POS and NER Model

• Word Neighbors Visualization via Fasttext Embeddings

• Building Annotation Job using Docaano

Additional Information

• https://soc.uofsa.edu

• Venkat Gudivada, Dean, School of Computing, venkat.gudivada@uofsa.edu

• Raju Chamarthi, Provost and Chief Academic Officer, raju@uofsa.edu

Questions